Spatial distributions, feeding ecologies, and behavioral interactions of four rabbitfish species (Siganus unimaculatus, S. virgatus, S. corallinus, and S. puellus)

Study species

Four rabbitfish species were examined in the present study (Siganus unimaculatus, S. virgatus, S. corallinus, and S. puellus; Figs. 1D–1G), all of which are commonly found in the Okinawan region (Masuda & Kobayashi, 1994).

Spatial distributions of the fishes

To clarify the size-specific spatial distributions of the four rabbitfish species, underwater visual observations were conducted at Sekisei Lagoon and Nagura Bay in the Yaeyama Islands, Okinawa, in the southern part of the East China Sea (Figs. 1A, 1B). To obtain a more generalized understanding of their spatial distributions, observations were conducted during two periods: between June 2016 and January 2017, and between June 2017 and February 2018. Both series of observations were carried out at 63 study sites with an inter-site distance of ca. 2 km, allowing almost the entire area of Sekisei Lagoon and Nagura Bay to be surveyed (Fig. 1C). At each site, a 20-minute underwater visual survey was conducted along a 5-m-wide transect between 0830 and 1600 h using SCUBA equipment, following the methods of Nanami et al. (2013). During each 20-minute survey, a portable global positioning system receiver was attached to a buoy that was towed, allowing the distance covered to be recorded. At each site, the number of individuals observed and their total length (TL) were recorded for each of the four rabbitfish species. The number of individuals was then converted into a density (number of individuals per 100 m ×5 m) using the distance data. The average distances covered were 343.6 ± 43.8 m (mean ±  standard deviation (SD)) and 370.6 ± 54.7 m for the first and second series of observations, respectively.

A diving computer (Xtender Quattro; SCUBAPRO) was used to record the depth profile at 30-second intervals and PCLogBook software (SCUBAPRO) was then used to download the depth profile to a personal computer. This yielded 40 depth values for each site (two depth points per minute ×20 min), which were averaged for analysis. The water depths ranged from 2.6 m to 12.4 m and from 3.2 m to 12.1 m in the first and second series of observations, respectively.

Spatial variation in substrate characteristics and depth at the study sites

To evaluate the substrate availability at each site, digital video images (moving pictures) of the substrate were recorded. Static images were then obtained at 10-second intervals using QuickTime Player Pro software (version 7.6), yielding 121 static images per 20-minute video image (from 0 to 1,200 s). The substrate at the center of each static image on the monitor of a personal computer was recorded for analysis. The substrate was divided into 16 categories for analysis, following Nanami et al. (2013) with some modification: (1) branching Acropora spp. (e.g., A. formosa), (2) tabular Acropora spp. (e.g., A. hyacinthus), (3) bottlebrush Acropora spp. (e.g., A. carduus), (4) branching corals except for Acropora spp. (e.g., branching Pocillopora spp., Montipora spp., and Porites spp.), (5) massive corals (e.g., massive Porites spp. and Faviidae), (6) other living corals (e.g., encrusting corals and leafy corals), (7) dead branching Acropora spp., (8) dead tabular Acropora spp., (9) dead bottlebrush Acropora spp., (10) dead branching corals except for Acropora spp., (11) dead other corals, (12) soft corals, (13) coral rubble, (14) rocks (calcium carbonate substratum with a lower substrate complexity), (15) sand, and (16) seaweed (e.g., Padina minor and Sargassum spp.). Both dead corals and rocks were covered by epilithic algae (Nanami, 2016).

Spatial distribution analysis

Individual fish were categorized into three size classes: small (TL ≤ 10 cm), medium (11 cm ≤ TL ≤ 15 cm), and large (TL ≥ 16 cm). The relationship between the spatial distribution of each size class for the four rabbitfish species and environmental characteristics (16 substrates and depth) was analyzed using canonical correspondence analysis (CCA) with CANOCO software (Ter Braak & Smilauer, 2002). This analysis was performed using averaged data from the two series of observations. To extract the environmental characteristics that significantly affected the spatial distribution, software options set for forward selection were applied.

Stomach contents analysis

To clarify the inter-specific differences in food items taken, the four rabbitfish species were sampled between February and August 2014 as follows: Siganus unimaculatus, n = 10, fork length (FL) = 149.0–166.0 mm, sampling period = February–July; S. virgatus, n = 10, FL = 159.5–215.5 mm, sampling period = June–July; S. corallinus, n = 10, FL = 126.0–205.5 mm, sampling period = April–August; and S. puellus, n = 10, FL = 148.0–219.0 mm, sampling period = April–August). Each individual was collected by spearing the head and was immediately placed in an icebox to minimize pain and any loss of stomach contents.

The stomach was dissected from each specimen and any food items in the stomach were recorded. The food items were sorted into 12 categories based on Hoey, Brandl & Bellwood (2013) with some modification: (1) brown foliose algae, (2) brown styloid algae (brown algae with a fine stalk-like shape), (3) brown filamentous algae, (4) other brown algae (brown algae that could not be sorted into the above-mentioned categories), (5) red foliose algae, (6) red styloid algae (red algae with a fine stalk-like shape), (7) red filamentous algae, (8) other red algae (red algae that could not be sorted into the above-mentioned categories), (9) green algae, (10) seagrass, (11) sponges, and (12) benthic invertebrates except for sponges (including Amphipoda, Anthozoa, Ascidiacea, Bivalvia, Bryozoa, Copepoda, Decapoda, Gastropoda, Hydrozoa, Isopoda, Ophiuroidea, Ostracoda, Polychaeta, and Radiolaria). Each food item was spread on a mesh sheet (mesh size = 5 mm ×5 mm) taking care not to overlay one item on another and the number of mesh intervals covered by each food item was counted. The number of mesh intervals covered was regarded as the volume of the food item.

The degree of similarity of food items among the 40 individuals (4 species ×10 individuals) was determined by calculating the Bray–Curtis similarity coefficient based on the percentage of each food item using PRIMER software (version 6). A dendrogram was then produced using the group-average linkage method.

Feeding behavior analysis

To examine inter-specific differences in feeding substrates, the feeding behaviors of the four species were observed around Ishigaki Island in December 2013 and data on the feeding substrates were also recorded. The estimated TL (cm), number of bites per 5 min (feeding rate), and number of bites during a continuous feeding period (foraging bouts; sensu (Bellwood & Choat, 1990) were counted using SCUBA equipment or snorkeling between 0900 and 1600 h. The numbers of individuals and TL ranges were as follows: S. unimaculatus, n = 17, TL = 16.0–20.0 cm; S. virgatus, n = 16, TL = 16.0–22.0 cm; S. corallinus, n = 11, TL = 19.0–26.0 cm; and S. puellus, n = 17, TL = 15.0–23.0 cm.

Preliminary observations revealed that the feeding substrates were almost completely limited to dead corals and rocks. Therefore, they were categorized according to structural features (i.e., concave, flat, or convex) rather than microhabitat (e.g., branching corals, massive corals, or tabular corals). The feeding substrates were classified into three categories, following Bellwood & Choat (1990): (1) concave substrate (e.g., inter-branch space of dead branching coral colonies and dents in rocks), (2) flat substrate (e.g., surface of dead tabular coral colonies and flat-shaped rocks), and (3) convex substrate (e.g., surface of dead massive coral colonies).

The feeding rates and numbers of foraging bouts on each of the three substrates among the four species were compared using one-way analysis of variance (ANOVA) followed by the post-hoc Games-Howell test.

Body shape analysis

To examine the relationship between feeding substrates and body shape characteristics, inter-specific body shape variations were analyzed using elliptic Fourier descriptors, which allow the shape of a closed two-dimensional contour to be analyzed (Iwata & Ukai, 2002) (see Furuta et al. (1995), Iwata et al. (1998), Iwata & Ukai (2002), and Nanami & Shimos (2013) for details). Briefly, the contour of the fish body shape was traced from an arbitrary point on the body surface around the body of the fish to return to the start point and was then projected on horizontal (x) and vertical (y) axes, allowing the contours along the x- and y-coordinates to be obtained. These contours were expressed as a waveform and approximated using the elliptic Fourier expansion. The body shape characteristics were summarized using principal component analysis (PCA). The estimated body shape was drawn against three values (−2 SD, the average, and +2 SD) of the principal component scores for PCA axes 1 and 2 using the obtained Fourier coefficients. This estimated body shape was then used to recognize the relationship between the two PCA axes and the body shape.

The analyses were conducted using SHAPE software version 1.3 (Iwata & Ukai, 2002). Ten individuals were analyzed for each species with the following FL ranges: S. unimaculatus, 147.0–188 mm; S. puellus, 148.0–195 mm; S. corallinus, 162.5–212.0 mm; and S. virgatus, 155.0–215.5 mm. The shape of the lateral aspect of each fish (excluding the dorsal, pelvic, anal, and caudal fins) was used in the analysis. To examine inter-specific differences in body shape characteristics, the PCA scores for PCA axes 1 and 2 were compared using one-way ANOVA followed by the post-hoc Games-Howell test for multiple comparisons among the four species.

Behavioral interactions

To examine the behavioral characteristics of the four rabbitfish species during inter- and intraspecific encounters, aggressive interactions were recorded at the study site by following pairs of fish by snorkeling or using SCUBA equipment. First, an observer approached a focal pair (two individuals swimming within 50 cm of each other with no aggressive behavior) or a solitary individual, taking care not to scare them (the distance between the focal pair and observer was always >3 m). The fish were then directly observed for 5–10 min. Whenever the focal pair (or individual) encountered other pairs or individuals, the observer recorded the species that was encountered and any occurrence of (1) an attack (chasing other pairs or individuals), (2) an agonistic display (approaching other pairs or individuals and showing fin displays, such as raising the dorsal, pelvic, and anal fins), and (3) no interaction (no noticeable change in behavior when pairs were within 50 cm of each other), following Nanami (2015). At the end of the observation period, a new focal pair or individual was chosen taking care not to choose the same pair or individual.

The frequency of occurrence of the three types of behaviors was calculated for each pair. For the purposes of analysis, “attack” and “agonistic display” were combined and regarded as “agonistic behavior,” giving two categories of inter- and intraspecific interactions: (1) agonistic behavior and (2) no interaction. The ratios of the two behavioral categories were compared using one-way ANOVA followed by post-hoc Scheffé tests for multiple comparisons among the four species.

Spatial distributions

Small individuals of all four species had a relatively limited distribution through the study area (e.g., the inner area of Nagura Bay and western area of Sekisei Lagoon), whereas large individuals had a more extensive distribution and medium-sized individuals had an intermediate distribution (Fig. 2).

CCA revealed that the 63 study sites could be divided into five groups based on the species- and size-specific spatial distributions (Figs. 3A–3C): group A, which included two sites with a higher density of small S. virgatus individuals; group B, which included one site with a higher density of small S. virgatus and S. unimaculatus individuals; group C, which included four sites with a higher density of small, medium, and large S. unimaculatus individuals; Group D, which included eight sites with a higher density of medium and large individuals of all four species; and Group E, which included 48 sites with a higher density of large individuals of all four species.

Group A consisted of relatively shallow sites with a greater coverage of branching Acropora spp. and seaweed, Group B consisted of sites with a greater coverage of branching Acropora spp., Group C consisted of relatively deep sites with a greater coverage of bottlebrush Acropora spp. and dead Acropora spp., and group E consisted of sites with a greater coverage of rocks (Fig. 3D). Overall, group A, B and C sites were mainly located in sheltered parts of the study area (the inner areas of Nagura Bay and Sekisei Lagoon), whereas group E sites were located in both exposed (outer edge of the study area) and sheltered regions (Fig. 3E).

Diets

Cluster analysis revealed that all 10 individuals of S. unimaculatus mainly fed on brown foliose algae, red foliose algae, and red styloid algae, with sponges and bryozoa also being found to some extent (Fig. 4, Table S1). By contrast, the main food item of S. virgatus was brown foliose algae. The main food items of S. corallinus varied between individuals. Overall, brown foliose algae and red styloid algae were taken but hydrozoans and green algae were also found to some extent (Fig. 4). According to the cluster analysis, the 10 S. corallinus individuals were divided into three groups: a group in which S. unimaculatus predominated (six individuals; group A in Fig. 4), a group in which S. virgatus predominated (three individuals; group B), and the remainder (one individual; group D). All 10 individuals of S. puellus mainly had sponges in their stomachs (Fig. 4) but bryozoa and hydrozoa were also taken to some extent (Table S1), indicating that this species had a very different diet from the other three species.

Feeding behaviors

The average feeding rates for S. unimaculatus, S. virgatus, S. corallinus, and S. puellus were 44.5, 35.5, 47.5, and 14.8 bites per 5 min, respectively (Fig. 5A). Siganus unimaculatus mainly used concave substrates whereas S. virgatus mainly used convex substrates for feeding (one-way ANOVA and Games-Howell test, p < 0.05). However, no significant differences in feeding substrate were found for S. corallinus and S. puellus (Fig. 5A).

The average numbers of foraging bouts for S. unimaculatus, S. virgatus, S. corallinus, and S. puellus were 2.7, 2.7, 3.6, and 1.5 bites/bout, respectively (Fig. 5B). The number of foraging bouts was significantly higher on concave substrates for S. unimaculatus and convex and flat substrates for S. virgatus (one-way ANOVA and Games-Howell test, p < 0.05). Again, there were no significant differences in the number of foraging bouts among substrates for S. corallinus and S. puellus (Fig. 5B).

Body shape characteristics

Overall, the body shapes of the fishes could be divided into three types: (1) slender body with a protruding snout (S. unimaculatus), (2) slender body with a low degree of snout protrusion (S. puellus), and (3) relatively deeper body with a low degree of snout protrusion (S. virgatus and S. corallinus) (Fig. 6).

The two-dimensional plot of the two PCA axes revealed that a positive value along PCA axis 1 indicated a deeper body and a low degree of snout protrusion, while a negative value indicated a shallower body and a protruding snout. These differences explained 69.7% of the observed body shape variation. Siganus virgatus had significantly higher scores for PCA axis 1 than the other three species, and S. corallinus had significantly higher scores than S. unimaculatus and S. puellus (S. virgatus > S. corallinus > S. unimaculatus and S. puellus; Games-Howell test, p < 0.05).

A positive value along PCA axis 2 indicated a low degree of snout protrusion, while a negative value indicated a protruding snout. These differences explained 14.3% of the observed body shape variation. Siganus puellus had significantly higher scores for PCA axis 2 than the other three species, whereas S. unimaculatus had significantly lower scores than the other three species (S. puellus > S. virgatus and S. corallinus  >  S. unimaculatus; Games-Howell test, p < 0.05).

Behavioral interactions

Siganus unimaculatus, S. virgatus, and S. puellus showed significantly greater frequencies of agonistic behaviors during conspecific encounters than heterospecific encounters (one-way ANOVA and Scheffé test, p < 0.0001 for all three species) (Fig. 7). Siganus corallinus also exhibited the same trend, although this was not significant.

Size-specific spatial distribution in relation to habitat characteristics

Nanami (2015) previously investigated the fine-scale size-specific spatial distribution of the rabbitfish species S. unimaculatus, with a focus on the spatial arrangement of its home ranges and the effect of body size on these. However, this study did not consider the relationship between the large-scale spatial distribution and environmental characteristics. Furthermore, although Hoey, Brandl & Bellwood (2013) conducted a detailed analysis of the spatial distributions of 11 rabbitfish species inhabiting the Great Barrier Reef, investigating differences between three shelf regions (inner-shelf, mid-shelf, and outer-shelf) and four habitats (back reef, reef flat, reef crest, and reef slope), they did not consider size-specific spatial variations in relation to substrate characteristics. In the present study, the size-specific spatial distributions of four rabbitfish species were examined over a scale of several to tens of kilometers on an Okinawan coral reef, representing the first such study in this region.

The findings of the present study suggested that inter-specific variation in the spatial distribution was higher for small individuals than for large individuals. It appears that sheltered areas with a greater coverage of branching and bottlebrush Acropora spp. may act as nursery grounds for the four rabbitfish species (sensu (Beck et al., 2001). The fine complexity of these coral species would provide a suitable habitat and refuge space for the fishes, likely reducing mortality by predation (Beukers & Jones, 1998; Almany, 2004; Nanami et al., 2013). By contrast, large individuals of all four species were found in rocky areas. Thus, habitat partitioning does not appear to be the main mechanism that allows the coexistence of these species.

Food item partitioning in relation to feeding behavior

The findings indicated that food item partitioning occurs among the four study species, with S. unimaculatus, S. virgatus, and S. puellus in particular showing clear differences in the food items taken without any evidence of habitat partitioning. Such food item partitioning among species of the same genus has also been found in other families (Nanami & Shimose, 2013; Brandl, William & Bellwood, 2015; Liedke et al., 2017).

It has previously been shown that body shape characteristics can affect the microhabitat that is used for foraging (Brandl, Hoey & Bellwood, 2014; Brandl, William & Bellwood, 2015; Brandl & Bellwood, 2016) and the food items that are taken (Carpenter, 1996; Nanami & Shimose, 2013). The present study demonstrated the species-specific use of feeding substrates among the four rabbitfish species. The findings also suggested that the slender body and protruding snout of S. unimaculatus lend themselves to use of a concave substrate, whereas the stouter body shape with low level of snout protrusion in S. virgatus meant that it rarely used a concave substrate for feeding. Siganus corallinus did not exhibit any substrate-specific use for feeding and took both brown foliose algae and red styloid algae, which was intermediate between S. unimaculatus and S. virgatus. Since S. corallinus has similar (albeit significantly different) body shape characteristics to S. virgatus, it may alter its main feeding substrate to reduce inter-specific competition, which may also explain why its main food items were intermediate between S. unimaculatus and S. virgatus.

Overall, these results suggest that S. unimaculatus and S. virgatus partition their feeding substrate according to their body shape characteristics, which results in them taking different main food items, enabling their coexistence without any substantial difference in spatial distribution. By contrast, the greater use of flat substrates by S. corallinus may enable its coexistence with S. unimaculatus and S. virgatus. A fine (i.e., species-level) taxonomic classification of food items may provide further detail around the mechanisms that allow these three species to coexist.

Siganus puellus mainly fed on sponges, suggesting that it could use all three substrates without experiencing any inter-specific competition. This species has a slender body with a low degree of snout protrusion, which may enable the use of all three substrates. Siganus puellus also exhibited a lower feeding rate and average number of foraging bouts than the other species, possibly due to the sponge density being lower than the algal density.

Another aspect that needs to be considered when examining inter-specific differences in food items is the ability to digest and assimilate algae (Clements & Choat, 1995). Therefore, since red and brown algae have been shown to have considerably different carbohydrate compositions (Stiger-Pouvreau, Bourgougnon & Deslandes, 2016), inter-specific differences in the ability to digest the respective food items among the four rabbitfish species should be investigated in the future.

Frequencies of agonistic behaviors

Nanami (2015) previously revealed that individuals of S. unimaculatus have distinct home ranges that are maintained by intraspecific interactions among conspecific pairs. However, few studies have shown the degree of inter- and intraspecific behavioral interactions among multiple rabbitfish species. In the present study, all four species exhibited a greater frequency of conspecific than heterospecific agonistic behavior, suggesting that intraspecific competition for food items and foraging substrates is greater than inter-specific competition for these resources. Overall, such intraspecific agonistic behavior would enhance the syntopic distribution of the four rabbitfish species.

Nanami (2015) showed that similar-sized conspecific pairs of S. unimaculatus did not have overlapping territories whereas no such pattern was found for dissimilar-sized conspecific pairs. However, no data on the spatial arrangement of home ranges are available for S. virgatus, S. corallinus, and S. puellus. Therefore, further studies are needed to clarify why conspecific agonistic behavior was frequently observed for all four rabbitfish species.

It should be noted that other herbivorous species that feed on macroalgae and turf algae (e.g., in the families Kyphosidae and Acanthuridae) may also compete with the rabbitfish species for food (Clements & Choat, 1995; Choat, Clements & Robbins, 2002). No encounters with non-rabbitfish herbivores (e.g., Kyphosus and Naso spp.) that are considered potential competitors were observed during the field observations, suggesting there may be spatial differences between these species. However, to comprehensively understand the mechanisms that allow the four rabbitfish species to coexist, more precise studies are needed on the spatial distributions and feeding ecologies of non-rabbitfish species.